OpenAI reveals its first AI processor: Jalapeño
TL;DR
OpenAI has introduced Jalapeño, its first custom AI server chip. Built with Broadcom, the ASIC is designed for inference: serving ChatGPT answers, running Codex-style agents, and processing live model requests. The chip is meant to power current and future large language models while reducing OpenAI's dependence on scarce Nvidia GPUs. Broadcom CEO Hock Tan claims it matches Nvidia Blackwell and Google's TPU performance.
Nauti's Take
This is less a chip launch than a power move in AI infrastructure. OpenAI does not want to stay permanently dependent on Nvidia while every new agent workflow consumes more compute.
The story is still PR-heavy, though: without independent benchmarks, Jalapeño is mostly a promise. The real test is whether it merely lowers costs or unlocks products that would be too expensive on standard GPUs.
Sources